how accurate is this ntp

[hans]

i believe very accurate or let's say extremely precise.
otherwise the calculation for the location of the lightning strokes could not be done.

of course i switched on the NTP server functionality. as with every watch one wants have it running very exactly. even if it is in a way of precision i never can use it. i configured my solaris server to use my system red as stratum 1. additional i configured 2 stratum 1 server near my location in vienna with "noselect". as i know from the past it is always bad if delay and especially jitter is high. and one additional from the location where the time is made ( braunschweig :-) also with "noselect". i also have a DCF77 which has per definition a big systemic error. i adjusted this with the "fudge" statement to be _near_ the GPS signal. when i look i have most of the time a situation like this:

PHP Code:

# ntpq -pn
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
+192.168.241.93  .DCFa.           1 u   62 1024  377    0.743   -0.706   4.345
*192.168.241.190 .GPS.            1 u  365 1024  377    0.917   -0.001   0.139
 192.53.103.108  .PTB.            1 u  418 1024  377   46.284   -8.819   0.673
 178.189.127.147 .ATOM.           1 u  996 1024  377   21.794   -5.281   0.512
 131.130.251.107 .PPS.            1 u  205 1024  377   18.983   -6.134  12.341 


my server takes the time from system red (*192.168.241.190) and has a typically offset between 6 and 8 ms against the others stratum 1. that means that system red is in average 7 ms in the future against the others. unfortunately the "fudge" statement works only for reference clocks and not for the system red i configured in my server which is stratum 2.

i don't know how many of you are using system red as ntp server. any feedback about experience from you about this topic is welcome.

maybe the future brings a possibility to enter an offset for the ntp server functionality. but i know it's not the primary function of this system.

finally i want to say thank you to Egon and others for this great project.

kind regards
hans

--

[Tobi]

The NTP server functionality was just a quick and dirty try and there's already a correction offset implemented. A submillisecond accuracy was not the goal so far. I don't know whether it would be really possible with the current (network-)hardware due to varying internal delays due to polling.

[hans]

Dear Tobi,

many thanks for your reply.
i can understand completely that it's not a goal to implement a NTP server. i don't think it was Q&D. maybe it was done quick but it is working very well. in my ntp.conf i have some other stratum 1 server ( some only for monitoring ) and after a time always the system red is taken as a reference. so it can't be so bad.
and if you address the 100 mbit interface this is completely enough for NTP at a home network.
i read there is already implemented a system correction offset. it may be not so complicated to implement an input field on the GUI where the user can enter his offset. his would make happy those users which like to play around with NTP too.

kind regards
hans

[Ross.Wheeler]

Pretty happy with how mine is performing...

Code:

# ntpq -p
     remote           refid      st t when poll reach   delay   offset  jitter
================================================================================
+ali-syd-1.*  192.65.182.4      3 u  464 1024  376   47.729   -0.511   0.857
-ali-syd-2.*  192.65.182.4      3 u  468 1024  376   45.720    2.297   1.902
+ali-syd-3.*  192.65.182.4      3 u  277 1024  376   46.019    0.193   2.098
-starone.*.   115.126.160.4     3 u  664 1024  377    0.658   -1.522   0.301
*sparks.*      .GPS.            1 u  843 1024  375    0.765    0.134   0.019      <-- system RED

[hans]

Hi Ross

that's interesting. many thanks for your reply.
if i calculate the average of your stratum 3 server's offset ( 0.114 ms ) it's almost identical as system red. so it fits perfect. and you are using "only" stratum 3 references. mine is not just a snapshot. i see similar values very often. my system red has an offset for about +7 ms. i have to write a script which calculates the average of hundreds of values out of the peerstats file. but i am sure this will bring a similar result. no idea why this behaves in that way.

which version are you running. i have
FirmwareRev. 7.4 / Aug 17 2014 16:41:16
HardwareRev. 10.4 (PCB-Id 2)

BTW: where did you configure the last line in the posting where your stations are listed ?

// hans

[Ross.Wheeler]

if i calculate the average of your stratum 3 server's offset ( 0.114 ms ) it's almost identical as system red. so it fits perfect. and you are using "only" stratum 3 references.

That was a display from one of my servers. It's configured to use several NTP servers - I have 8 servers in different locations up to 1000km apart, in a mesh. They use some external servers, themselves, and my BO controller.

BTW: where did you configure the last line in the posting where your stations are listed ?

So all I've done is slightly obfuscate the hostnames, the rest of the data is exactly what ntpq gave me when I queried it.

which version are you running. i have
FirmwareRev. 7.4 / Aug 17 2014 16:41:16
HardwareRev. 10.4 (PCB-Id 2)

I should update, but for now all 4 of my systems are stable and happy...
FirmwareRev. 7.3 / Aug 16 2014 21:22:37
HardwareRev. 10.3 (PCB-Id 1)

[hans]

Hi Ross

Thanks, maybe the different firmware version makes the different.

>> BTW: where did you configure the last line in the posting where your stations are listed ?
> So all I've done is slightly obfuscate the hostnames, the rest of the data is exactly what ntpq gave me when I queried it.

You misunderstood. Maybe I didn't ask exactly. I mean this line:
Stations: 812, 848, 849, 852
It seems this is added automatically. Everybody who add a reply or new posting has listed his station(s). But not for me. So I ask how to configure that.

// hans

[Tobi]

Hans, please log in here: http://www.blitzortung.org/Webpages/inde...=en&page=4

Then switch to the forum tab, so the forum can see that you are a participant.

[hans]

Hi Tobi,

I did so. Let's see if this works.

// Hans

---

No, it doesn't work.
But I think I found the issue.
Within the link you sent I am "hans.mayer" with user id 1143 and station id 1269.
Here in the forum I am user "hans". This doesn't match.
Maybe I should register as "hans.mayer" in the forum.
I think this happened because in the forum I am much longer than my station is registered.

// Hans

[Tobi]

@Hans: Yes, please use your other username. We are getting really off-topic here, so for further questions you should create a separate thread or ask via email.

[Knickohr]

NTP works very well, the best timesource ever, and it's in under my control ;-)

I use it for all my PC, router, cams, ... specially for a meteor detection camera. Offset is only a ms max.

Thomas

[hans.mayer]

hi Thomas,

thanks for your reply. a question, how did you measure the offset less than 1 ms ?

i don't say NTP works bad. it's the opposite, it is great. and i am happy to have it.
what i tried to ask is the fact, if somebody else had the same experience as me, that this NTP has a little offset against others stratum 1 time servers. actually it's not be worth to speak about it. but i did. ( some things i want to know exactly - sorry )

as i wrote in my first posting i configured two additional stratum 1 server as reference.
now i wrote a small shell/awk/gnuplot script for visualization of these endless NTP log files.
attached you can find 2 jpg's
it shows the offset to a stratum 1 in Vienna and the other is in Braunschweig.
it's calculated with the yesterday's statistic files.
my first guess with 6 ms was not so bad. ( in my case )


kind regards
hans